Design And Implementation Of A Multi-Agent Based Dynamic Scheduling System

With the entry of 21st century and the coming of global manufacturing age, manufacturing environments being more dynamic and competitive, market demand and environment/social pressures require modern manufacturing systems to adapt themselves to become more agile, intelligent, and integrated systems. Dynamic scheduling in manufacture, understood as the problem of suitable assignment of manufacturing resources to tasks/jobs within a specified time window and coping with a set of constraints, was a major problem in manufacturing research and development. So in the new situation in manufacturing environments, there are needs to develop new methods and new mechanisms to solve the agile production-scheduling problem.The Agent paradigm represents one of the most prominent approaches to build complex and flexible intelligent systems including agile production-scheduling system. Agent has a certain degree of autonomy to reason about and to make decisions by itself, has the capability to interact with other Agents, and has the capability to independently solve a part of the global problem. Therefore Agent paradigm and Agent-based system give a new way to solve dynamic scheduling in modern manufacturing.In this paper, a Multi-Agent based Scheduling System (MASS) is presents to solve the danymic scheduling problem and a prototype system for Agent-based dynamic scheduling in manufacturing environments has build up. The main works of this paper are as follows:Firstly, the model of a heterarchical Agent-based job-shop dynamic scheduling system, constructed by Manager Agent, Part Agent and Recourse Agent, is built up. Furthermore, inner models for these three kinds of Agents are given to meet the requirements of decision-making and communication among Agents.Secondly, "Rule-based negotiation" scheduling mechanism and strategies, which integrated rule-based scheduling methods and the negotiation mechanism among Agents, are proposed to create a proper way for mutual cooperation among Agents, and to improve theintelligence, autonomy and cooperative ability of Agents. The proposed strategies realized the flexibility of distributed system and the adaptive ability for Agents in dynamic situation.Thirdly, the prototype system based upon the above Agent-based dynamic scheduling system for job shop is developed. With Java and Objective-oriented technology, models for intelligent Agents objects and other non-intelligent objects, such as tasks and operations in scheduling, as well as the communication models are built up to construct a simulating test bed for multi-Agent based scheduling system.At the end, the simulations applying rule-based negotiation strategies on the normal and abnormal matters (equipment failure etc.) in job shop are made on the developed test bed. Simulation experiments which applying some kinds of classical scheduling strategies into the proposed Agent-based system are also made.The simulation results have successfully shown that the approach can effectively handle the factual problems, including abnormal situations in dynamic job-shop scheduling; it has the capability to generate feasible schedulers to meet the demands of modem manufacturing.